Lossless Compression of Double-Precision Floating-Point Data for Numerical Simulations: Highly Parallelizable Algorithms for GPU Computing